An Adaptive RAM-DFE for Storage Channels

A modification of the decision feedback equalizer (DFE) is presented and analyzed for use in channels with trailing nonlinear intersymbol interference, most notably binary saturation-recording channels. We call this modified DFE the RAM-DFE because a look-up table replaces the transversal filter feedback section of the DFE, and this table can be easily implemented with a random access memory. The feedforward section of the equalizer remains linear. A general nonlinear Markov (or finite-state machine) model is used to model the nonlinear intersymbol interference (ISI) channel. With this Markov model, we introduce a method for computing the minimum-mean-squared-error settings of the feedforward filter coefficients, and also the feedback filter and look-up table contents for the linear DFE and the RAM-DFE, respectively. The RAM-DFE with these settings can yield significant performance improvement over the linear DFE for channels with trailing nonlinear ISI. We also introduce and analyze globally convergent gradient-type algorithms for updating the feedforward section coefficients and the contents of the feedback table. This improvement is verified on data taken from a representative thin-film disk system, along with several other channels. © 1991 IEEE